LeCun vs. Manning clash over whether LLMs can understand the physical world

ShuiwangJi · x · 2026-10-02

A public spat over whether language models can understand the physical world: Yann LeCun argues no, while Chris Manning fires back that "that is wrong and has been proven to be wrong." Stanford professor Shuiwang Ji weighed in agreeing with LeCun, quipping that "when legends argue, move aside."

The exchange captures a core academic fault line: whether knowledge learned from language constitutes genuine understanding of the physical world or just statistical correlation — a debate over the fundamental limits of LLMs.

Related event: LeCun and Manning clash over whether LLMs understand the physical world(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →